Predict the Banana 🍌
Surprise: every banana is very slightly radioactive. It's packed with potassium, and a tiny sliver of that is potassium-40 (K-40) — an atom that randomly "pops" (decays) now and then. You can never predict which atom pops next. But you can predict the crowd astonishingly well. That gap is today's big idea.
Random atoms. Predictable patterns.
🍌 The Banana Counter
First, guess: in one real banana, how many K-40 atoms do you think pop in a whole minute? Type a guess, then run the counter and watch the random sparks.
Your guess
Run the clock
A banana pops about 15 times every second — roughly 900 times a minute. Yet it never runs out: it holds around 800,000,000,000,000,000 K-40 atoms, and each one only decays once every billion-or-so years on average. The rate depends only on the nucleus of potassium-40 — not on whether the banana is fresh, ripe, or rotten.
Each spark is a surprise…
You can't say which atom will pop, or exactly when. Every single spark is genuinely random — there's no countdown, no schedule.
The count lands close every time
Run it again and the number of pops barely changes. Random events, piled up by the millions, make a steady, predictable rate.
⚖️ One atom vs. a million
Real K-40 is way too slow to watch a single atom. So for this one, pretend each atom flips a coin every second — heads, it decays. Press the button and compare the two sides.
The single atom flips wildly between YES and NO — unknowable. The million always lands near half. Same coins, totally different predictability.
🎯 Quick quiz: which can you predict?
Tap the ones whose result you could predict pretty well before it happens. Then check your answers.
📊 Many runs: the crowd settles down
Pick a group size, then run it over and over. Each run lands a little differently — but watch how a bigger group hugs the middle far more tightly. (Each bar = how often a run gave that percentage.)
Group size
Why? This is the law of large numbers: the bigger the crowd, the smaller the wobble around the average. Ten atoms is chaos; a hundred thousand is a tight, predictable spike — even though every single atom is still flipping a fair coin.
Random ≠ unknowable
We cannot predict which atom pops, or when. But across a huge group, the overall pattern is rock-solid. Randomness at the bottom can build order at the top.
That's why this sits next to Predict the AI and the chaos tools: different roads — chaos, randomness — leading to the same question of what we can predict and at what scale.
Try these out loud
Why can we predict the group when we can't predict the individual? Where else does this show up — coin flips, raindrops on a roof, votes, website clicks, traffic?
And: would a riper banana pop faster? (No — same nucleus, same rate. Ripeness is chemistry; decay is nuclear.)
For teachers & grown-ups
Big idea: radioactive decay is random for an individual atom but statistical for large groups — a clean, real-world gateway to probability and the law of large numbers. Important correction: the decay rate is a fixed property of the potassium-40 nucleus; it does not depend on the banana's age, ripeness, or condition. Real figures: a banana contains roughly 8×10¹⁷ K-40 atoms and emits about 15 decays per second (~15 Bq); K-40's half-life is ~1.25 billion years, so depletion over a human lifetime is utterly negligible (the "Banana Counter" sparks are real-rate but the dots are symbolic — the banana never measurably runs low). The "one atom vs a million" and "many runs" demos switch to a sped-up coin-flip model (p = ½) purely so a single atom does something visible; the lesson — relative fluctuation shrinks like 1/√N — is identical for true exponential decay, where counts in a fixed window are Poisson-distributed. Connections: this pairs naturally with the chaos tools (two different origins of unpredictability — sensitive dependence vs. genuine randomness) and reinforces that "deterministic," "random," and "predictable" are independent ideas.